# Import Library (Satu-satunya sel kode dalam template ini)
import numpy as np
import matplotlib.pyplot as plt
import librosa
import soundfile as sf
from PIL import Image
import cv2
from IPython.display import Audio, HTML, display
import os

# Set matplotlib untuk menampilkan plot inline
%matplotlib inline

# Tampilkan versi library untuk dokumentasi
print("Library versions:")
print(f"NumPy: {np.__version__}")
print(f"Matplotlib: {plt.matplotlib.__version__}")
print(f"Librosa: {librosa.__version__}")
print(f"OpenCV: {cv2.__version__}")

# Tambahkan import lain jika diperlukan saat mengerjakan tugas

Library versions:
NumPy: 2.2.5
Matplotlib: 3.10.6
Librosa: 0.11.0
OpenCV: 4.10.0

# Muat file audio dan tampilkan metadata dasar
audio_path = os.path.join('audio', 'audio1.wav')

# Membaca audio (y: array amplitudo, sr: sample rate)
y, sr = librosa.load(audio_path, sr=None, mono=False)  # mono=False agar bisa deteksi jumlah kanal

# Jika y 1D: mono, jika 2D: stereo/multi-channel
if y.ndim == 1:
  num_channels = 1
  channel_info = 'Mono'
  total_samples = y.shape[0]
else:
  num_channels = y.shape[0]
  channel_info = f'Stereo ({num_channels} channels)' if num_channels == 2 else f'{num_channels} channels'
  total_samples = y.shape[1]

duration_sec = total_samples / sr

print(f"Path file      : {audio_path}")
print(f"Sample rate    : {sr} Hz")
print(f"Durasi         : {duration_sec:.2f} detik")
print(f"Jumlah kanal   : {channel_info}")
print(f"Total sampel   : {total_samples}")

Path file      : audio\audio1.wav
Sample rate    : 44100 Hz
Durasi         : 25.07 detik
Jumlah kanal   : Mono
Total sampel   : 1105448

plt.figure(figsize=(12, 4))
librosa.display.waveshow(y, sr=sr, color='blue', x_axis='time')
plt.title('Waveform (Bentuk Gelombang)')
plt.xlabel('Waktu (detik)')
plt.ylabel('Amplitudo')
plt.grid(True)
plt.tight_layout()
plt.show()

C:\Users\Naufal Haris\AppData\Local\Temp\ipykernel_7924\3691184900.py:2: FutureWarning: waveshow() keyword argument 'x_axis' has been renamed to 'axis' in version 0.10.0.
	This alias will be removed in version 1.0.
  librosa.display.waveshow(y, sr=sr, color='blue', x_axis='time')

# 3. Membuat Visualisasi
fig, ax = plt.subplots(nrows=2, ncols=1, sharex=True, figsize=(12, 10))
fig.suptitle('Analisis Visual Sinyal Audio', fontsize=16)

# b. Menampilkan Spectrogram
# Lakukan Short-Time Fourier Transform (STFT)
D = librosa.stft(y)
# Konversi amplitudo ke desibel (dB)
S_db = librosa.amplitude_to_db(np.abs(D), ref=np.max)
img_spec = librosa.display.specshow(S_db, sr=sr, x_axis='time', y_axis='log', ax=ax[0])
ax[0].set_title('Spectrogram')
ax[0].set_ylabel('Frekuensi (Hz)')
fig.colorbar(img_spec, ax=ax[0], format='%+2.0f dB', label='Intensitas')


# c. Menampilkan Mel-spectrogram
# Hitung Mel-spectrogram
M = librosa.feature.melspectrogram(y=y, sr=sr)
# Konversi ke desibel (dB)
M_db = librosa.power_to_db(M, ref=np.max)
img_mel = librosa.display.specshow(M_db, sr=sr, x_axis='time', y_axis='mel', ax=ax[1])
ax[1].set_title('Mel-spectrogram')
ax[1].set_xlabel('Waktu (detik)')
ax[1].set_ylabel('Frekuensi (Mel)')
fig.colorbar(img_mel, ax=ax[1], format='%+2.0f dB', label='Intensitas')


# Merapikan layout dan menampilkan plot
plt.tight_layout(rect=[0, 0.03, 1, 0.95])
plt.show()

import librosa
import soundfile as sf # Pustaka untuk menyimpan file audio
# Muat file audio dan tampilkan metadata dasar
audio_path = os.path.join('audio', 'audio1.wav')

# Membaca audio (y: array amplitudo, sr: sample rate)
y_original, sr_original = librosa.load(audio_path, sr=None, mono=False)  # mono=False agar bisa deteksi jumlah kanal

print(f"Audio Asli sebelum resampling:")
print(f"Laju Sampel Asli: {sr_original} Hz")
print(f"Durasi Asli: {librosa.get_duration(y=y_original, sr=sr_original):.2f} detik")

# melakukan down sampling ke kualitas telepon

# Downsample ke kualitas telepon (8,000 Hz)
y_telepon, sr_telepon = librosa.load(audio_path, sr=8000)

print("\nAudio setelah Resampling:")
print(f"Laju Sampel Telepon: {sr_telepon} Hz")

print(f"\n--- Analisis Durasi ---")
print(f"Durasi Asli (di {sr_original} Hz): {librosa.get_duration(y=y_original, sr=sr_original):.2f} detik")
print(f"Durasi Telepon (di {sr_telepon} Hz): {librosa.get_duration(y=y_telepon, sr=sr_telepon):.2f} detik")

Audio Asli sebelum resampling:
Laju Sampel Asli: 44100 Hz
Durasi Asli: 25.07 detik

Audio setelah Resampling:
Laju Sampel Telepon: 8000 Hz

--- Analisis Durasi ---
Durasi Asli (di 44100 Hz): 25.07 detik
Durasi Telepon (di 8000 Hz): 25.07 detik

import matplotlib.pyplot as plt
import numpy as np

# Buat spectrogram untuk audio kualitas telepon
D_telepon = librosa.stft(y_telepon)
S_db_telepon = librosa.amplitude_to_db(np.abs(D_telepon), ref=np.max)

# Tampilkan
fig, ax = plt.subplots(figsize=(10, 4))
img = librosa.display.specshow(S_db_telepon, sr=sr_telepon, x_axis='time', y_axis='log', ax=ax)
fig.colorbar(img, ax=ax, format='%+2.0f dB')
ax.set_title(f'Spectrogram pada Laju Sampel {sr_telepon} Hz')
plt.show()

# Simpan file audio yang sudah di-resample
sf.write('audio_telepon_8000Hz.wav', y_telepon, sr_telepon)

print("\nFile audio yang telah di-resample berhasil disimpan!")

File audio yang telah di-resample berhasil disimpan!

import numpy as np
import matplotlib.pyplot as plt
from scipy.io import wavfile
from scipy import signal

# Baca file audio dari folder audio
fs, audio = wavfile.read('audio/audio2.wav')

# Informasi audio
print(f"\n📊 INFORMASI AUDIO:")
print(f"   - Sample rate: {fs} Hz")
print(f"   - Durasi: {len(audio)/fs:.2f} detik")
print(f"   - Jumlah sample: {len(audio)}")
print(f"   - Channels: Mono")

# Visualisasi
fig = plt.figure(figsize=(15, 12))

# 2. Spektrogram
ax2 = plt.subplot(4, 1, 2)
f, t, Sxx = signal.spectrogram(audio, fs, nperseg=1024)
pcm = ax2.pcolormesh(t, f, 10 * np.log10(Sxx + 1e-10), shading='gouraud', cmap='viridis')
ax2.set_ylabel('Frequency (Hz)', fontsize=11)
ax2.set_xlabel('Time (s)', fontsize=11)
ax2.set_title('Original Audio Spectrogram', fontweight='bold', fontsize=13)
ax2.set_ylim(0, 10000)  # Focus pada 0-10kHz
plt.colorbar(pcm, ax=ax2, label='Power (dB)')

plt.tight_layout()
plt.savefig('01_original_audio_analysis.png', dpi=150, bbox_inches='tight')
plt.show()

📊 INFORMASI AUDIO:
   - Sample rate: 44100 Hz
   - Durasi: 8.08 detik
   - Jumlah sample: 356352
   - Channels: Mono

# Fungsi untuk membuat filter
def apply_filter(audio, fs, filter_type, cutoff):
    nyquist = fs / 2
    
    if filter_type == 'lowpass':
        b, a = signal.butter(4, cutoff / nyquist, btype='low')
    elif filter_type == 'highpass':
        b, a = signal.butter(4, cutoff / nyquist, btype='high')
    elif filter_type == 'bandpass':
        b, a = signal.butter(4, [cutoff[0] / nyquist, cutoff[1] / nyquist], btype='band')
    
    filtered = signal.filtfilt(b, a, audio)
    return filtered

# Eksperimen dengan berbagai cutoff frequency
cutoff_values = [500, 1000, 2000]

# ============================================
# VISUALISASI SEMUA FILTER
# ============================================

fig, axes = plt.subplots(10, 1, figsize=(14, 20))

# Plot audio original
print("Plotting original audio...")
f_orig, t_orig, Sxx_orig = signal.spectrogram(audio, fs, nperseg=1024)
axes[0].pcolormesh(t_orig, f_orig, 10 * np.log10(Sxx_orig + 1e-10), shading='gouraud', cmap='viridis')
axes[0].set_ylabel('Frequency (Hz)', fontsize=10)
axes[0].set_title('Original Audio Spectrogram', fontweight='bold', fontsize=12, pad=10)
axes[0].set_ylim(0, 10000)

idx = 1

# LOW-PASS FILTERS
print("\nLOW-PASS FILTERS:")
for cutoff in cutoff_values:
    print(f"   ✓ Testing low-pass {cutoff} Hz...")
    filtered = apply_filter(audio, fs, 'lowpass', cutoff)
    f, t, Sxx = signal.spectrogram(filtered, fs, nperseg=1024)
    axes[idx].pcolormesh(t, f, 10 * np.log10(Sxx + 1e-10), shading='gouraud', cmap='viridis')
    axes[idx].set_ylabel('Frequency (Hz)', fontsize=10)
    axes[idx].set_title(f'Low-pass Filter - Cutoff {cutoff} Hz', fontweight='bold', fontsize=11)
    axes[idx].set_ylim(0, 10000)
    
    # Tambahkan garis cutoff
    axes[idx].axhline(y=cutoff, color='red', linestyle='--', linewidth=1.5, alpha=0.6)
    
    idx += 1

# HIGH-PASS FILTERS
print("\nHIGH-PASS FILTERS:")
for cutoff in cutoff_values:
    print(f"   ✓ Testing high-pass {cutoff} Hz...")
    filtered = apply_filter(audio, fs, 'highpass', cutoff)
    f, t, Sxx = signal.spectrogram(filtered, fs, nperseg=1024)
    axes[idx].pcolormesh(t, f, 10 * np.log10(Sxx + 1e-10), shading='gouraud', cmap='viridis')
    axes[idx].set_ylabel('Frequency (Hz)', fontsize=10)
    axes[idx].set_title(f'High-pass Filter - Cutoff {cutoff} Hz', fontweight='bold', fontsize=11)
    axes[idx].set_ylim(0, 10000)
    
    # Tambahkan garis cutoff
    axes[idx].axhline(y=cutoff, color='red', linestyle='--', linewidth=1.5, alpha=0.6)
    
    idx += 1

# BAND-PASS FILTERS
print("\nBAND-PASS FILTERS:")
bandpass_ranges = [(300, 3000), (500, 3500), (1000, 4000)]
for low, high in bandpass_ranges:
    print(f"   ✓ Testing band-pass {low}-{high} Hz...")
    filtered = apply_filter(audio, fs, 'bandpass', [low, high])
    f, t, Sxx = signal.spectrogram(filtered, fs, nperseg=1024)
    axes[idx].pcolormesh(t, f, 10 * np.log10(Sxx + 1e-10), shading='gouraud', cmap='viridis')
    axes[idx].set_ylabel('Frequency (Hz)', fontsize=10)
    axes[idx].set_title(f'Band-pass Filter - {low}-{high} Hz', fontweight='bold', fontsize=11)
    axes[idx].set_ylim(0, 10000)
    
    # Tambahkan garis cutoff bawah dan atas
    axes[idx].axhline(y=low, color='red', linestyle='--', linewidth=1.5, alpha=0.6)
    axes[idx].axhline(y=high, color='orange', linestyle='--', linewidth=1.5, alpha=0.6)
    
    idx += 1

axes[-1].set_xlabel('Time (s)', fontsize=11)
plt.tight_layout()
plt.savefig('02_all_filters_comparison.png', dpi=150, bbox_inches='tight')
print(f"\n💾 Visualisasi disimpan: 02_all_filters_comparison.png")
plt.show()

Plotting original audio...

LOW-PASS FILTERS:
   ✓ Testing low-pass 500 Hz...
   ✓ Testing low-pass 1000 Hz...
   ✓ Testing low-pass 2000 Hz...

HIGH-PASS FILTERS:
   ✓ Testing high-pass 500 Hz...
   ✓ Testing high-pass 1000 Hz...
   ✓ Testing high-pass 2000 Hz...

BAND-PASS FILTERS:
   ✓ Testing band-pass 300-3000 Hz...
   ✓ Testing band-pass 500-3500 Hz...
   ✓ Testing band-pass 1000-4000 Hz...

💾 Visualisasi disimpan: 02_all_filters_comparison.png

print("\n📊 Membuat visualisasi perbandingan before-after...")

# --- PILIH FILTER TERBAIK UNTUK PERBANDINGAN ---
# Untuk noise kipas, band-pass paling efektif.
# Kita pilih rentang 300 Hz - 3000 Hz untuk mengisolasi suara vokal.
best_low_cutoff = 300
best_high_cutoff = 3000
filtered_audio = apply_filter(audio, fs, 'bandpass', [best_low_cutoff, best_high_cutoff])

# --- MEMBUAT PLOT 2x2 ---
fig, axes = plt.subplots(2, figsize=(16, 10), gridspec_kw={'height_ratios': [1, 1]})
fig.suptitle(f'Perbandingan Audio: Sebelum vs Sesudah Filter Band-Pass ({best_low_cutoff}-{best_high_cutoff} Hz)', fontsize=16, fontweight='bold')

# --- BARIS ATAS: AUDIO ORIGINAL ---

# 2. Original Spectrogram (Kanan Atas)
f_orig, t_orig, Sxx_orig = signal.spectrogram(audio, fs, nperseg=1024)
pcm1 = axes[0].pcolormesh(t_orig, f_orig, 10 * np.log10(Sxx_orig + 1e-10), shading='gouraud', cmap='magma')
axes[0].set_title('Original Audio - Spectrogram', fontweight='bold')
axes[0].set_ylabel('Frequency (Hz)')
axes[0].set_ylim(0, 8000) # Batasi Y-axis agar lebih fokus
plt.colorbar(pcm1, ax=axes[0], label='Power (dB)')


# --- BARIS BAWAH: AUDIO SETELAH FILTER ---

# 4. Filtered Spectrogram (Kanan Bawah)
f_filt, t_filt, Sxx_filt = signal.spectrogram(filtered_audio, fs, nperseg=1024)
pcm2 = axes[1].pcolormesh(t_filt, f_filt, 10 * np.log10(Sxx_filt + 1e-10), shading='gouraud', cmap='magma')
axes[1].set_title(f'Filtered Audio - Spectrogram', fontweight='bold')
axes[1].set_xlabel('Time (s)')
axes[1].set_ylabel('Frequency (Hz)')
axes[1].set_ylim(0, 8000)
# Tandai area yang difilter
axes[1].axhline(y=best_low_cutoff, color='cyan', linestyle='--', linewidth=2, alpha=0.8, label=f'Low Cutoff {best_low_cutoff} Hz')
axes[1].axhline(y=best_high_cutoff, color='cyan', linestyle='--', linewidth=2, alpha=0.8, label=f'High Cutoff {best_high_cutoff} Hz')
axes[1].legend(fontsize='small')
plt.colorbar(pcm2, ax=axes[1], label='Power (dB)')


plt.tight_layout(rect=[0, 0, 1, 0.96]) # Atur layout agar suptitle tidak tumpang tindih
plt.savefig('03_before_after_comparison.png', dpi=150)
print(f"💾 Perbandingan disimpan sebagai: 03_before_after_comparison.png")
plt.show()

📊 Membuat visualisasi perbandingan before-after...

C:\Users\Naufal Haris\AppData\Local\Temp\ipykernel_1808\776772587.py:41: UserWarning: Creating legend with loc="best" can be slow with large amounts of data.
  plt.tight_layout(rect=[0, 0, 1, 0.96]) # Atur layout agar suptitle tidak tumpang tindih

💾 Perbandingan disimpan sebagai: 03_before_after_comparison.png

import librosa
import librosa.display
import matplotlib.pyplot as plt
import numpy as np
import soundfile as sf
import os

# --- 1. Persiapan dan Pemuatan Audio ---
# Tentukan path ke file audio
audio_path = os.path.join('audio', 'audio1.wav')

# Cek apakah file ada
if not os.path.exists(audio_path):
    print(f"Error: File tidak ditemukan di '{audio_path}'")
    exit()

print("Memuat file audio...")
# Muat file audio dengan librosa
# y = time series audio, sr = sample rate
y, sr = librosa.load(audio_path)
print("File audio berhasil dimuat.")

# --- 2. Visualisasi Audio Asli ---
print("Membuat visualisasi audio asli...")
plt.figure(figsize=(14, 10))

# Waveform Asli
plt.subplot(3, 2, 1)
librosa.display.waveshow(y, sr=sr)
plt.title('Waveform Audio Asli')
plt.xlabel('Waktu (detik)')
plt.ylabel('Amplitudo')

# Spectrogram Asli
plt.subplot(3, 2, 2)
D_original = librosa.stft(y)  # Short-Time Fourier Transform
S_db_original = librosa.amplitude_to_db(np.abs(D_original), ref=np.max)
librosa.display.specshow(S_db_original, sr=sr, x_axis='time', y_axis='log')
plt.title('Spectrogram Audio Asli')
plt.xlabel('Waktu (detik)')
plt.ylabel('Frekuensi (Hz)')
plt.colorbar(format='%+2.0f dB')

# --- 3. Proses Pitch Shifting ---
print("Melakukan pitch shifting...")
# Pitch shifting sebanyak +7 semitone (perfect fifth)
y_shifted_7 = librosa.effects.pitch_shift(y, sr=sr, n_steps=7)

# Pitch shifting sebanyak +12 semitone (satu oktaf penuh)
y_shifted_12 = librosa.effects.pitch_shift(y, sr=sr, n_steps=12)
print("Pitch shifting selesai.")

# --- 4. Visualisasi Audio Setelah Pitch Shifting (+12) ---
print("Membuat visualisasi audio setelah pitch shifting...")
# Waveform Setelah Pitch Shift (+12)
plt.subplot(3, 2, 3)
librosa.display.waveshow(y_shifted_12, sr=sr)
plt.title('Waveform Setelah Pitch Shift (+12 Semitone)')
plt.xlabel('Waktu (detik)')
plt.ylabel('Amplitudo')

# Spectrogram Setelah Pitch Shift (+12)
plt.subplot(3, 2, 4)
D_shifted_12 = librosa.stft(y_shifted_12)
S_db_shifted_12 = librosa.amplitude_to_db(np.abs(D_shifted_12), ref=np.max)
librosa.display.specshow(S_db_shifted_12, sr=sr, x_axis='time', y_axis='log')
plt.title('Spectrogram Setelah Pitch Shift (+12 Semitone)')
plt.xlabel('Waktu (detik)')
plt.ylabel('Frekuensi (Hz)')
plt.colorbar(format='%+2.0f dB')

# Tampilkan semua plot
plt.tight_layout()
plt.show()

Memuat file audio...
File audio berhasil dimuat.
Membuat visualisasi audio asli...
Melakukan pitch shifting...
Pitch shifting selesai.
Membuat visualisasi audio setelah pitch shifting...

# --- 4. Penggabungan dan Penyimpanan Audio ---
print("Menggabungkan audio yang telah di-pitch shift...")
# Gabungkan kedua audio secara berurutan
y_combined = np.concatenate((y_shifted_7, y_shifted_12))

# Simpan audio gabungan ke file baru
output_path = 'audio_chipmunk_combined.wav'
sf.write(output_path, y_combined, sr)
print(f"Audio gabungan berhasil disimpan sebagai '{output_path}'")


# --- 5. Visualisasi Audio Gabungan (FIGURE 2) ---
# Membuat figure kedua khusus untuk audio gabungan
plt.figure(figsize=(16, 5))
librosa.display.waveshow(y_combined, sr=sr)
plt.title('Waveform Audio Gabungan (25 detik pertama: +7, 25 detik kedua: +12 Semitone)', fontsize=14)
plt.axvline(x=25, color='r', linestyle='--', label='Batas Pitch Shift (+7 -> +12)')
plt.xlabel('Waktu (detik)')
plt.ylabel('Amplitudo')
plt.legend()
plt.grid(True, alpha=0.5)


# --- 6. Tampilkan Semua Plot ---
plt.show()

Menggabungkan audio yang telah di-pitch shift...
Audio gabungan berhasil disimpan sebagai 'audio_chipmunk_combined.wav'

import librosa
import librosa.display
import matplotlib.pyplot as plt
import numpy as np
import soundfile as sf
import os
import pyloudnorm as pyln
from scipy.signal import butter, lfilter

# --- 1. Persiapan dan Pemuatan Audio ---
# Tentukan path ke file audio
audio_path = os.path.join('audio', 'audio1.wav')

# Cek apakah file ada
if not os.path.exists(audio_path):
    print(f"Error: File tidak ditemukan di '{audio_path}'")
    # Membuat file audio dummy jika tidak ada untuk pengujian
    print("Membuat file audio dummy 'audio/audio1.wav' untuk demonstrasi...")
    sr_dummy = 22050
    duration = 25
    y_dummy = np.array([])
    for i in range(5):
        amplitude = (i + 1) / 5.0
        t = np.linspace(0., 5., int(5 * sr_dummy), endpoint=False)
        y_segment = amplitude * np.sin(2. * np.pi * (220 + i*50) * t)
        y_dummy = np.concatenate((y_dummy, y_segment))
    
    os.makedirs('audio', exist_ok=True)
    sf.write(audio_path, y_dummy, sr_dummy)
    y, sr = y_dummy, sr_dummy
else:
    print("Memuat file audio...")
    y, sr = librosa.load(audio_path)
    print("File audio berhasil dimuat.")


# --- 2. Proses Pitch Shifting ---
print("Melakukan pitch shifting...")
y_shifted_7 = librosa.effects.pitch_shift(y=y, sr=sr, n_steps=7)
y_shifted_12 = librosa.effects.pitch_shift(y=y, sr=sr, n_steps=12)
print("Pitch shifting selesai.")


# --- 3. Penggabungan Audio ---
print("Menggabungkan audio yang telah di-pitch shift...")
y_combined = np.concatenate((y_shifted_7, y_shifted_12))


# --- 4. Audio Processing Chain ---
print("\nMemulai Audio Processing Chain...")
y_processed = y_combined.copy()

# 4.1. Silence Trimming
print("1. Menghapus keheningan (Silence Trimming)...")
y_processed, _ = librosa.effects.trim(y_processed, top_db=20)

# 4.2. Equalizer (Simple High-pass Filter)
def high_pass_filter(data, cutoff, fs, order=5):
    nyq = 0.5 * fs
    normal_cutoff = cutoff / nyq
    b, a = butter(order, normal_cutoff, btype='high', analog=False)
    y = lfilter(b, a, data)
    return y

print("2. Menerapkan Equalizer (High-pass filter at 80Hz)...")
y_processed = high_pass_filter(y_processed, cutoff=80, fs=sr)

# 4.3. Compression
print("3. Menerapkan Kompresi Dinamis...")
# Parameter ini akan 'menekan' sinyal yang lebih keras, mengurangi rentang dinamis

def dynamic_compress(y, threshold_dB=-20, ratio=4):
    y_abs = np.abs(y)
    y_dB = 20 * np.log10(y_abs + 1e-10)
    mask = y_dB > threshold_dB
    gain_reduction = np.where(mask, (y_dB - threshold_dB) / ratio, 0.0)
    gain = 10 ** (-gain_reduction / 20)
    return y * gain

y_processed = dynamic_compress(y_processed, threshold_dB=-20, ratio=4)

# 4.4. Noise Gate (Implementasi Sederhana)
print("4. Menerapkan Noise Gate...")
frame_length = 2048
hop_length = 512
rms = librosa.feature.rms(y=y_processed, frame_length=frame_length, hop_length=hop_length)[0]
gate_threshold = 0.01 
for i, r in enumerate(rms):
    if r < gate_threshold:
        start_sample = i * hop_length
        end_sample = start_sample + hop_length
        y_processed[start_sample:end_sample] = 0

# 4.5. Gain/Fade
print("5. Menambahkan Fade In dan Fade Out...")
fade_len_samples = int(sr * 1.0) # 1 detik fade
fade_in = np.linspace(0, 1, fade_len_samples)
fade_out = np.linspace(1, 0, fade_len_samples)
y_processed[:fade_len_samples] *= fade_in
y_processed[-fade_len_samples:] *= fade_out

# Simpan audio SEBELUM normalisasi untuk perbandingan
y_before_norm = y_processed.copy()

# 4.6. Loudness Normalization
print("6. Melakukan Normalisasi Loudness ke -16 LUFS...")
target_lufs = -16.0
meter = pyln.Meter(sr) # Buat meter
# Ukur loudness
loudness = meter.integrated_loudness(y_processed)
# Hitung gain yang diperlukan dan terapkan
gain_db = target_lufs - loudness
y_processed = pyln.normalize.loudness(y_processed, loudness, target_lufs)
print(f"   - Loudness Awal: {loudness:.2f} LUFS")
print(f"   - Loudness Target: {target_lufs:.2f} LUFS")
print(f"   - Gain diterapkan: {gain_db:.2f} dB")

y_after_norm = y_processed

# Simpan audio hasil akhir
output_path_processed = 'audio_chipmunk_processed.wav'
sf.write(output_path_processed, y_after_norm, sr)
print(f"\nAudio yang telah diproses sepenuhnya disimpan sebagai '{output_path_processed}'")


# --- 5. Visualisasi ---

# FIGURE 1: Perbandingan Audio Asli vs Pitch Shifted (Sama seperti sebelumnya)
plt.figure(figsize=(16, 12))
plt.suptitle('Perbandingan Audio Asli vs Pitch Shifted', fontsize=20, y=0.95)
# ... (Kode plot asli, shift +7, shift +12 tetap sama)
# Plot Audio Asli
plt.subplot(3, 2, 1); librosa.display.waveshow(y, sr=sr); plt.title('Waveform Audio Asli')
plt.subplot(3, 2, 2); librosa.display.specshow(librosa.amplitude_to_db(np.abs(librosa.stft(y)), ref=np.max), sr=sr, x_axis='time', y_axis='log'); plt.title('Spectrogram Audio Asli'); plt.colorbar(format='%+2.0f dB')
# Plot Audio Setelah Pitch Shift +7
plt.subplot(3, 2, 3); librosa.display.waveshow(y_shifted_7, sr=sr); plt.title('Waveform Setelah Pitch Shift (+7)')
plt.subplot(3, 2, 4); librosa.display.specshow(librosa.amplitude_to_db(np.abs(librosa.stft(y_shifted_7)), ref=np.max), sr=sr, x_axis='time', y_axis='log'); plt.title('Spectrogram Setelah Pitch Shift (+7)'); plt.colorbar(format='%+2.0f dB')
# Plot Audio Setelah Pitch Shift +12
plt.subplot(3, 2, 5); librosa.display.waveshow(y_shifted_12, sr=sr); plt.title('Waveform Setelah Pitch Shift (+12)')
plt.subplot(3, 2, 6); librosa.display.specshow(librosa.amplitude_to_db(np.abs(librosa.stft(y_shifted_12)), ref=np.max), sr=sr, x_axis='time', y_axis='log'); plt.title('Spectrogram Setelah Pitch Shift (+12)'); plt.colorbar(format='%+2.0f dB')
plt.tight_layout(rect=[0, 0, 1, 0.93])


# FIGURE 2: Perbandingan SEBELUM dan SESUDAH Normalisasi
plt.figure(figsize=(16, 8))
plt.suptitle('Perbandingan Sebelum vs Sesudah Normalisasi Loudness (-16 LUFS)', fontsize=20, y=0.98)

# Waveform Sebelum Normalisasi
plt.subplot(2, 2, 1)
librosa.display.waveshow(y_before_norm, sr=sr)
plt.title('Waveform SEBELUM Normalisasi')
plt.xlabel('Waktu (detik)')
plt.ylabel('Amplitudo')
plt.ylim(-1, 1)

# Spectrogram Sebelum Normalisasi
plt.subplot(2, 2, 2)
D_before = librosa.stft(y_before_norm)
S_db_before = librosa.amplitude_to_db(np.abs(D_before), ref=np.max)
librosa.display.specshow(S_db_before, sr=sr, x_axis='time', y_axis='log')
plt.title('Spectrogram SEBELUM Normalisasi')
plt.colorbar(format='%+2.0f dB')

# Waveform Sesudah Normalisasi
plt.subplot(2, 2, 3)
librosa.display.waveshow(y_after_norm, sr=sr, color='orange')
plt.title('Waveform SESUDAH Normalisasi')
plt.xlabel('Waktu (detik)')
plt.ylabel('Amplitudo')
plt.ylim(-1, 1)

# Spectrogram Sesudah Normalisasi
plt.subplot(2, 2, 4)
D_after = librosa.stft(y_after_norm)
S_db_after = librosa.amplitude_to_db(np.abs(D_after), ref=np.max)
librosa.display.specshow(S_db_after, sr=sr, x_axis='time', y_axis='log')
plt.title('Spectrogram SESUDAH Normalisasi')
plt.colorbar(format='%+2.0f dB')
plt.tight_layout(rect=[0, 0, 1, 0.93])


# FIGURE 3: Waveform Hasil Akhir
plt.figure(figsize=(16, 5))
librosa.display.waveshow(y_after_norm, sr=sr, color='green')
plt.title('Waveform Hasil Akhir Setelah Seluruh Proses', fontsize=14)
plt.xlabel('Waktu (detik)')
plt.ylabel('Amplitudo')
plt.grid(True, alpha=0.5)

# Tampilkan semua plot
plt.show()

Memuat file audio...
File audio berhasil dimuat.
Melakukan pitch shifting...
Pitch shifting selesai.
Menggabungkan audio yang telah di-pitch shift...

Memulai Audio Processing Chain...
1. Menghapus keheningan (Silence Trimming)...
2. Menerapkan Equalizer (High-pass filter at 80Hz)...
3. Menerapkan Kompresi Dinamis...
4. Menerapkan Noise Gate...
5. Menambahkan Fade In dan Fade Out...
6. Melakukan Normalisasi Loudness ke -16 LUFS...

c:\Users\Naufal Haris\.conda\envs\multimedia_env\lib\site-packages\pyloudnorm\normalize.py:62: UserWarning: Possible clipped samples in output.
  warnings.warn("Possible clipped samples in output.")

   - Loudness Awal: -35.24 LUFS
   - Loudness Target: -16.00 LUFS
   - Gain diterapkan: 19.24 dB

Audio yang telah diproses sepenuhnya disimpan sebagai 'audio_chipmunk_processed.wav'

import librosa
import librosa.display
import numpy as np
import soundfile as sf
import matplotlib.pyplot as plt

def analyze_audio(file_path):
    """Menganalisis file audio untuk mendapatkan tempo (BPM) dan kunci (key)."""
    # Muat file audio
    y, sr = librosa.load(file_path)

    # 1. Deteksi Tempo (BPM)
    tempo, _ = librosa.beat.beat_track(y=y, sr=sr)

    # 2. Estimasi Kunci (Key)
    chromagram = librosa.feature.chroma_stft(y=y, sr=sr)
    # Rata-ratakan nilai chroma di sepanjang waktu
    mean_chroma = np.mean(chromagram, axis=1)
    # Temukan pitch class dengan energi tertinggi
    dominant_pitch_class = np.argmax(mean_chroma)
    # Konversi ke nama not
    keys = ['C', 'C#', 'D', 'D#', 'E', 'F', 'F#', 'G', 'G#', 'A', 'A#', 'B']
    estimated_key = keys[dominant_pitch_class]

    return y, sr, tempo, estimated_key

# --- Langkah 1: Analisis Lagu Asli ---
# Ganti dengan path file lagu Anda
path_happier = 'audio/happier.wav'
path_sugar = 'audio/sugar.wav'

# Muat dan analisis kedua lagu
y_happier, sr_happier, tempo_happier, key_happier = analyze_audio(path_happier)
y_sugar, sr_sugar, tempo_sugar, key_sugar = analyze_audio(path_sugar)

print(f"Analisis 'Happier':")
print(f"  - Tempo: {tempo_happier[0]:.2f} BPM")
print(f"  - Estimasi Kunci: {key_happier}")
print("-" * 20)
print(f"Analisis 'Sugar':")
print(f"  - Tempo: {float(tempo_happier[0]):.2f} BPM")
print(f"  - Estimasi Kunci: {key_sugar}")
print("-" * 20)

# --- Langkah 2: Proses Remix (Stretching & Pitch Shifting) ---

# Tentukan target tempo dan kunci (kita samakan ke 'Sugar')
target_tempo = tempo_sugar
target_key_index = ['C', 'C#', 'D', 'D#', 'E', 'F', 'F#', 'G', 'G#', 'A', 'A#', 'B'].index('C#') # Kunci Sugar adalah Db/C#

# 1. Stretch: Samakan tempo 'Happier' ke tempo 'Sugar'
rate_stretch = float(target_tempo / tempo_happier[0])  # Ambil elemen pertama & konversi ke float
y_happier_stretched = librosa.effects.time_stretch(y_happier, rate=rate_stretch)

# 2. Pitch Shift: Samakan kunci 'Happier' ke kunci 'Sugar'
# Hitung perbedaan semitone yang dibutuhkan
current_key_index = ['C', 'C#', 'D', 'D#', 'E', 'F', 'F#', 'G', 'G#', 'A', 'A#', 'B'].index(key_happier)
semitones_to_shift = target_key_index - current_key_index
# Lakukan pitch shift
y_happier_pitched = librosa.effects.pitch_shift(y_happier_stretched, sr=sr_happier, n_steps=semitones_to_shift)

# --- Langkah 3: Gabungkan Lagu dengan Crossfading Sederhana ---

# Pastikan kedua audio memiliki panjang yang sama untuk penggabungan sederhana
len_target = min(len(y_happier_pitched), len(y_sugar))
y_happier_final = y_happier_pitched[:len_target]
y_sugar_final = y_sugar[:len_target]

# Buat efek crossfade sederhana (linear fade)
fade_duration = int(sr_happier * 5) # Crossfade selama 5 detik
fade_in = np.linspace(0., 1., fade_duration)
fade_out = np.linspace(1., 0., fade_duration)

# Terapkan fade di tengah-tengah audio
mid_point = len_target // 2
start_fade = mid_point - (fade_duration // 2)
end_fade = mid_point + (fade_duration // 2)

# Buat track baru untuk hasil remix
y_remix = np.zeros(len_target)
y_remix[:start_fade] = y_happier_final[:start_fade] # Bagian awal hanya Happier
y_remix[end_fade:] = y_sugar_final[end_fade:]     # Bagian akhir hanya Sugar

# Bagian crossfade
y_remix[start_fade:end_fade] = (y_happier_final[start_fade:end_fade] * fade_out) + (y_sugar_final[start_fade:end_fade] * fade_in)


# Simpan hasil remix ke file baru
output_filename = 'remix_happier_sugar.wav'
sf.write(output_filename, y_remix, sr_happier)
print(f"Remix berhasil disimpan sebagai '{output_filename}'")

# --- Langkah 4: Visualisasi Waveform dan Spectrogram Hasil Remix ---

fig, ax = plt.subplots(nrows=2, ncols=1, sharex=True, figsize=(12, 8))

# Tampilkan Waveform
librosa.display.waveshow(y_remix, sr=sr_happier, ax=ax[0], color='b')
ax[0].set_title('Waveform Hasil Remix')
ax[0].set_ylabel('Amplitudo')

# Tampilkan Spectrogram
D = librosa.stft(y_remix)
DB = librosa.amplitude_to_db(np.abs(D), ref=np.max)
librosa.display.specshow(DB, sr=sr_happier, x_axis='time', y_axis='log', ax=ax[1])
ax[1].set_title('Spectrogram Hasil Remix')
ax[1].set_ylabel('Frekuensi (Hz)')
ax[1].set_xlabel('Waktu (s)')

fig.tight_layout()
plt.show()

Analisis 'Happier':
  - Tempo: 89.10 BPM
  - Estimasi Kunci: C
--------------------
Analisis 'Sugar':
  - Tempo: 89.10 BPM
  - Estimasi Kunci: C#
--------------------

C:\Users\Naufal Haris\AppData\Local\Temp\ipykernel_1808\3747153256.py:52: DeprecationWarning: Conversion of an array with ndim > 0 to a scalar is deprecated, and will error in future. Ensure you extract a single element from your array before performing this operation. (Deprecated NumPy 1.25.)
  rate_stretch = float(target_tempo / tempo_happier[0])  # Ambil elemen pertama & konversi ke float

Remix berhasil disimpan sebagai 'remix_happier_sugar.wav'

Tugas Hands-on Pemrosesan Audio¶

Deskripsi Tugas¶

Soal 1: Rekaman dan Analisis Suara Multi-Level¶

Import Library¶

Load data audio1 dan exploration¶

Visualisasi dengan Waveform dan Spectogram¶

Analisis Waveform¶

Analisis spectogram¶

Resampling¶

memvisualisasikan hasil sampling¶

Analisis perbandingan hasil sampling dengan suara asli¶

Soal 2: Noise Reduction dengan Filtering¶

Load data dan visualisasi¶

Analisis Spectogram Noise¶

Filtering noise¶

Analisis Perbandingan¶

Comparison¶

PENJELASAN¶

Soal 3: Pitch Shifting dan Audio Manipulation¶

PENJELASAN¶

Perbandingan pitch¶

Soal 4: Audio Processing Chain¶

PENJELASAN¶

Soal 5: Music Analysis dan Remix¶

PENJELASAN¶

ANALISIS HASIL REMIX¶